大管径给排水管道径向变形自动检测

Automatic Detection of Radial Deformation of Large Diameter Water Supply and Drainage Pipelines

针对检测精度、召回率以及查全率较低的问题,本文提出一种大管径给排水管道径向变形自动检测方法。采用Mills交叉布列技术扫描管道,运用小波变换对扫描得到的图像进行去噪。根据去噪结果对大管径给排水管道径向变形区域进行粗检测,并对粗检测节点的预测区域进行分割,同时采用带空洞卷积的空间金字塔池化增强图像的感知能力,以有效实现大管径给排水管道径向变形自动检测。实验结果表明,所提方法能够有效提升变形自动检测精度、召回率以及查全率,获取较为满意的自动检测结果。

Aiming at the problems of low detection accuracy, recall rate and recall rate, an automatic detection method for radial deformation of large diameter water supply and drainage pipeline is proposed in this paper. Mills cross arrangement technology is used to scan the pipeline, and wavelet transform is used to denoise the scanned image. According to the denoising results, the radial deformation region of the large diameter water supply and drainage pipeline is roughly detected, and the prediction area of the coarse detection node is segmented. At the same time, the spatial pyramid pooling with hole convolution is used to enhance the image perception ability, so as to effectively realize the automatic detection of the radial deformation of large-diameter water supply and drainage pipeline. Experimental results show that the proposed method can effectively improve the accuracy, recall rate and recall rate of automatic deformation detection, and obtain satisfactory automatic detection results.